Deputy Prime Minister and Minister of Science and ICT Baek Kyung-hun emphasized the need to develop advanced frontier models to fundamentally address security threats posed by artificial intelligence (AI).
During a presidential briefing held at the Blue House on July 16, Baek stated, "To fundamentally respond to security issues arising from AI, it is necessary to develop not only security-specific models but also advanced frontier models."
President Yoon Suk Yeol highlighted the importance of securing independent AI capabilities, noting that foreign AI services, such as Anthropic's Mithos, are beyond domestic control and could be restricted or blocked at any time.
The president also urged for the strengthening of AI security systems, stating, "Security must be approached with the understanding that it can always be breached," and directed the establishment of a continuous assessment system that does not solely rely on companies' self-checks.
In response, Baek mentioned that the current project for an independent AI foundation model is underway, with plans to incorporate additional security data to develop a security-specific model by the end of the year. However, he stressed that to fundamentally address security threats, there is a need for an advanced frontier model equipped with comprehensive reasoning and analytical capabilities beyond specialized models.
President Yoon also instructed the creation of a system that can assess security vulnerabilities in a manner similar to actual attacks, without the consent of companies and public institutions.
Choi Woo-hyuk, head of the Information Security Network Policy Division at the Ministry of Science and ICT, explained that they are currently assessing services from over 2,000 companies under the ministry's jurisdiction with their consent and are conducting pilot projects with ten private and public institutions.
He added, "Under the current system, there is no legal basis to infiltrate systems without consent to check for vulnerabilities," and noted that after the pilot projects conclude, they are preparing related systems and legal foundations to allow for vulnerability assessments from the perspective of actual attackers without prior consent.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.